If we don't do the Europe project, what would you like instead ?

It’s perhaps time to consider alternatives if the Europe project doesn’t show enough support.

Here are some of the other projects that we can do instead:

1- ETF constituent data so that you can create universes based on constituents. (history since 2010 I think). Other use cases t.b.d
2- Integration of a precise earnings calendar integrated to screening , backtesting, stock pages, account pages
3- More broker integration (Fidelity, Ameritrade, Etrade, Alpaca)
4- Daily fundamental data for backtests instead of the current weekly snapshots

Thanks for your feedback & support

PS. #2 is in progress. we already have the data from Benzinga

Japan/ China data

In priority order:
1- Make TRADE work with non-us stocks.
2- More broker integration (Fidelity, Ameritrade, Etrade, Alpaca)
3- Daily fundamental data for backtests instead of the current weekly snapshots

Hi Marco, Earnings calendar for upcoming earnings is would be nice also, although it seems alot of sites don’t have good calendars and use estimated reporting dates which are often wrong. As earnings approach I’m often googling each company I own to find their press release indicating info on the actual upcoming earnings announcement, and whether it’s pre-market or post-market. The WeeksToQ field is useful, but actual confirmed upcoming earnings dates (including pre or post mkt) would be a very nice add imho for those of us who like to read the earnings releases and follow the conference calls.

Also, just thinking out loud here – I don’t know if glass door privides a csv of their company employee ratings, but I wonder if it would be useful from an factor perspective - especially in tech companies. When reading them they feel like useful information, but it’s hard to know.

My suggested order is as you posted + an option flag.

1- ETF constituent data so that you can create universes based on constituents. (history since 2010 I think). Other use cases t.b.d
2- Integration of a precise earnings calendar .to screening , backtesting, stock pages, account pages
3- More broker integration (Fidelity, Ameritrade, Etrade, Alpaca)
4- Daily fundamental data for backtests instead of the current weekly snapshots

5- Flag to identify optionable stocks.

I’m running ETF constituent data manually now, so it would be nice to have that as a feature.

My suggested order:

1- ETF constituent data so that you can create universes based on constituents. (history since 2010 I think). Other use cases t.b.d
2- Integration of a precise earnings calendar .to screening , backtesting, stock pages, account pages
3- Daily fundamental data for backtests instead of the current weekly snapshots
4- More broker integration (Fidelity, Ameritrade, Etrade, Alpaca)

I like #1-4 but would like #5 even more! Thumbs up

None of those are very important. By far the highest impact would be allowing the construction of long/short portfolios in simulations. E.g. long top 10% short bottom 10% based on ranking.

Marco,

I would like for P123 to just look into the cost of getting FactSet’s PIT version of the earnings estimates.

Probably not practical. “We looked into the details and it is not practical” would be enough for me. But the world is full of surprises. Maybe there is a holiday discount or something. You can’t be sure until you check.

Frankly, having a bunch of different data sources, all of questionable value (depending on one’s thoughts about the data and how much it might deviate from being perfectly PIT), is not as good as having one solid data source that I can bet my financial future upon with confidence.

I am willing to bet on just one stock each day if I have an edge and I know someone else is not getting the data ahead of me. Quality over quantity.

Some information about this here: ACCURATELY BACKTESTING FINANCIAL MODELS THROUGH POINT-IN-TIME CONSENSUS ESTIMATES

My personal question to myself: When I use XGBoost how will I ever know that I have found a real signal that I will be able to act upon in real-time in real-life?

I am only mentioning this because P123 asked and I truly believe it would be good for P123 if it turns out to be practical. I won’t bring it up as much when I know there is nothing to be done and “it is what it is.”

Thanks, truly, for your consideration of this.

Jim

I would also like to have formula weighting for books. That was something P123 wanted to introduce a while ago. There was even a non-functioning option on the simulated books web page to do this - (now removed).

Should be fairly easy to do.

  1. long/short portfolios in simulations
  2. backtesting based on intraday data

1 - API for selected pages would be great (currently having to scrape pages)

See feature request here → https://www.portfolio123.com/mvnforum/viewthread_thread,11555#!#66578

Thank you

Jerome

Data for Asian stocks.

IMHO it’s more important than European data. Asia it’s the future, Europe it’s the past.

Thanks.

I have to say I am quite fine with the functionality of p123!!!

On the top of my head:

  • I second backtesting based on intraday price data (e.g. sell if price above upper bollinger band etc.)

  • realtime buy limit orders on strats and ports (f.e. buy of a position only takes place if price is below 5% (configurable) of trigger price, stock gets only in port if limit gets hit (intraday would be nice!), if limit price gets not hit, port or strat stays in cash for this position, take the best 10 ranked stocks
    and put them into the limit price bucket. First limit that gets hit puts the stock in the port, then the other 10 best ranked stocks get discarded…)

  • chart pattern recocnition (especially breakout of a flat base or a flag)

I agree with Andreas in that I’m pretty happy with the current product/functionality. I think greater integration with the major brokers would bring the most bang for the P123 man hour buck. Seems to be a lot of consolidation in the Fintwit industry (with more to come) and with zero transaction fees seem to be a huge increase in trading among the public. This could be low hanging fruit that attracts the broadest customer base (at least on the retail side).

I would be careful with that request. It would require orders of magnitude more data, processing requirements, and probably cost. Intraday trading also puts you in direct competition with HFTs and I know who is going to win that battle. Quantopian already found that out.

I’m discussing this with an AI data scientist today as to what the best approach to this would be. I’m leaning towards image recognition , with perhaps some transformations before it’s submitted to the model for training/inference. Is there another way ? There are companies that do chart patterns and sell the data to brokers and such, but not sure what technology they used. My guess is something else since they have been around for a while . I rather develop the tech ourselves with AI.

*** Would also be an interesting community project since training a model for a chart patterns involves finding many by hand (thousands ?)

What kind of patterns would you like to see?

I’m no expert but I’d like to be able to find stocks that are consolidating after a large run-up. Not easy to do with rules.

What other chart patterns would you like to be able to screen for (and backtest of course) ?

Agree 100%. Tremendous effort and you’d be competing with Goldman Sachs who has proprietary semiconductor chips to gain an advantage

So I definitely agree with Steve on this.

But wouldn’t it be nice if we could start—at the open on the stock market—on the same footing as the HFTs even if we cannot hope to keep up during the day? With the same information on earnings estimates? Or at least backtest situations where they have an advantage and discard any models that do not work in that situation?

Or maybe others have a better explanation as to why the ports cannot keep up with the sims (i.e., sims are not PIT and/or we do not have the latest information for the ports).

Information on this from FactSet:
ACCURATELY BACKTESTING FINANCIAL MODELS THROUGH POINT-IN-TIME CONSENSUS ESTIMATES

Jim

Thank you.

Would not mind seeing that.

But me personally, I will specialize in XGBoost, maybe branching to TensorFlow occasionally WITH THE BEST DATA POSSIBLE INCLUDING EARNINGS ESTIMATES DATA. And consider using this once a got a firm handle on XGBoost.

And in truth I think one could just specialize in XGBoost forever—moving to alternative data if available. Or different Python programs for ETFs. Just me though. Like I said I would not mind seeing that.

Jim